369

Simple E1 Software Synchronization Algorithm Based on the Windows PIPE Concept. Yuriy Dorozhovets, Mykhailo Klymash

Abstract – This work demonstrate algorithm invented for FAS, and frame not containing the frame alignment signal - software realization at Windows PC, and demonstrate principles NFAS) are used for sending some signaling/management of E1 stream synchronization handling particular for fractional information, normally not used for any content data E1 to. transmission. Synchronization signal present in the bits 2…8 Keywords – E1, G.704, ITU-T, algorithm, PIPE, in the each FAS frame, format of this signal is in binary synchronization. 0011011(LSB). I. INTRODUCTION Every telecommunication system requires some low level data transmitting protocol. In Microwave radio relay, Satellite system is very common used E1 Frame to convey any transmitted data. E1 can handle with stream up to 2048kbps and it’s still a base for GSM Abis interface, HDLC and DCME links. E1 is based in few concepts: Fig.1 Binary FAS in the real E1 stream. · Frame alignment signal(FAS) All other features described in the G.704, like MFAS · 8000 fps CRC4 or 4kbprs are not the part of this work, because it’s · Timeslot optional E1 features. Data traffic can be conveyed inside E1 in any order but So base purpose of the realized software algorithm is FAS maximum throughput is up to 1984kbps. search/monitoring. There is wary important notation that proposed technic is designed to work in very hard condition, During projecting some receive equipment for E1 streams, it is where synchronization of the signal can be lost many times very important to implement simple and efficient E1 during receiving signal, and it’s very important to understand synchronization algorithm, and synchronization is only a half of all that output of such kind system will be very complex to future story. Most important to receiver site is a second part – information recovery. But without such algorithms there will resynchronization. This term mean recovering of the Frame be no any possibility to provide any data receiving. alignment signal in various points of the time. To provide this possibility in real-time mode it is required some cycle buffer. III. WINDOWS PIPE CONCEPT Almost all modern OS has very convenient way to do this – PIPE. For simple FIFO realization it is enough anonymous PIPE. There is two kinds of this objects – named and anonymous. It’s requiring less overhead than named pipes and easily to Named pipe is an extension to the traditional pipe concept on use. To create anonymous PIPE call: UNIX and Unix-like systems, and is one of the methods of inter- process communication. The concept is also found in Microsoft BOOL WINAPI CreatePipe( Windows, although the semantics differ substantially. A traditional __out PHANDLE hReadPipe, pipe is "unnamed" because it exists anonymously and persists only __out PHANDLE hWritePipe, for as long as the process is running. A named pipe is system- __in_opt LPSECURITY_ATTRIBUTES lpPipeAttributes, persistent and exists beyond the life of the process and must be __in DWORD nSize); deleted once it is no longer being used. Processes generally attach hReadPipe and hWritePipe – HANDLE for FIFO to the named pipes (usually appearing as a file) to perform inter- read/write functions. nSize – FIFO buffer size. process communication (IPC). As you see from this definition, the lpPipeAttributes – to set binary/massage mode, by default – main purpose of this concept is to organize multithread and binary. multiprocessors interaction, but also you can use PIPEs as very efficient FIFO buffers managed by OS. In this article we will demonstrate how to use it in this mode to provide software E1 synchronization and resynchronization module. II. E1 BASE SYNCHRONIZATION PRINCIPLES In accordance to ITU-T G.704(E1) bits 1 to 8 of the both frame types (frame containing the frame alignment signal -

Yuriy Dororzhovets, Mykhailo Klymash - Lviv Polytechnic National University , S. Bandery Str., 12, Lviv, 79013, UKRAINE, E-mail: [email protected] Fig.2 Base PIPE read algorithm.

TCSET'2012, February 21–24, 2012, Lviv-Slavske, Ukraine

Lviv Polytechnic National University Institutional Repository http://ena.lp.edu.ua 370

. When working with this concept bandwidthit’s very The obtained results are shown on the Fig. 4 and Fig. 5: important to know that starting from Windows XP it is required to control PIPE read side, because Microsoft make some undocumented changes in to PIPE logic and according to this changes all read function must control read size by algorithm. Another PIPE feature – blocking (by default), require read and write counters to avoid thread blocking during read from empty PIPE. IV. ALGORITHM DESCRIPTION Signal monitoring requires two threads, one realize write to FIFO function, another read and provide search/monitoring logic. When we have some E1 input, read write thread will call function what provide write frames to our FIFO buffer, Fig. 4 Output rate in different sync loss ps level. size of this buffer must be enough to compensate time wasted by search logic. In the first cycle read thread provide read 256 bytes (8 frames (4 FAS frames)) from FIFO using base PIPE read algorithm and search for 3 times FAS signal code sequence much in period from 2 to 32 bytes by step in 2 bytes, this way is very important to avoid wrong sequence detection in fractional E1. After this program will obtain E1 period and copy all good at least 6 good frames to output FIFO buffer. If period not detected drop all data, read new portion, and start from the beginning.

Fig.5 Random sync loss ps level. All research where provide in correct limits. This mean that to high level of the sync loss per second is uninteresting, because data recovery at these conditions can’t take place. VI. CONCLUSION This work described a simple E1 synchronization algorithm designed for software implementation at Windows PC using PIPE concept. After offered results of the performance research and demonstrated no linear dependence Fig. 3 E1 synchronization algorithm. of the algorithm performance and FAS loos frequency. Output Next depending to the input parameter (MismatchCounter) stream rate, and quality various directly proportional. program will monitor stream for detection of sequent Maximum algorithm performance in condition from 1 to 5 MismatchCounter loss of the FAS. To perform better algorithm FAS loos per second is more than 140Mbps, and this is flexibility proposed to use byte weight table, to avoid wrong equivalent to 70 parallel E1. So, proposed algorithm provides starting of the FAS search module. If detected loss of the FAS then opportunity to create software processing units with good call search module to detect new position of the FAS sequence and performance. loop monitoring again. All good frames during this process must be copied to output FIFO and read from it each 64 bytes (one FAS VII. REFERENCE + one NFAS frame). MismatchCounter in accordance to ITU-T recommendation is from 2 to 4, but it can be lager for some types [1] ITU-T G.704 Synchronous frame structures used at 1544, of traffic. 6312, 2048, 8448 and 44 736 kbit/s hierarchical levels 10.1998. [Electronic resource] http://www.itu.int/rec/T- V. ALGORITHM PERFORMANCE RESEARCH. REC-G.704/en It’s performed a research of output E1 speed limit [2] Microsoft developers network [Electronic resource] according to input stream synchronization loss. Research http://msdn.microsoft.com consist of the two parts, first linear sync loss and second – [3] ITU-T G.706 frame alignment and cyclicre dundancy random sync loss. All research provided at Windows PC check (CRC) procedures relating to basic frame structures based on Intel Core i3 3GHz processor.

TCSET'2012, February 21–24, 2012, Lviv-Slavske, Ukraine

Lviv Polytechnic National University Institutional Repository http://ena.lp.edu.ua